Felixor is an AI product studio. We design, build, and ship products where the intelligence is the foundation, not a feature.
We are the first kind. Felixor holds equity in what it builds, carries the roadmap, and lives with the consequences of its own design decisions. Some products we build alone. Some we build with a partner who brings the domain and the distribution.
What does not change is the starting point. We only build where a model can now do something it genuinely could not do three years ago, and where that capability removes work rather than adding a panel to it.
Most products begin with a market and go looking for features. We begin with something models can now do reliably, then hunt for the place that capability collapses an entire workflow instead of decorating one.
A working product in front of real users in weeks. Not a deck, not a pilot, not a design partner agreement. If the capability does not survive contact with real inputs, we would rather know in week three than in quarter three.
AI products break differently than software does. They fail confidently. We build the evaluation harness, the fallbacks, and the moments of honest uncertainty before we build the happy path, because those are the product.
Getting someone to try an AI product is easy right now. Getting them back on Wednesday is the entire game. We instrument for return, not for signups, and we keep building until that number moves.
Four products in active development. Each one started as a capability we could not stop thinking about.
Decision intelligence for operating teams.
View productAn operations copilot for work that happens in the physical world.
View productTurns what an organization knows into something it can be asked.
View productCash-flow intelligence for businesses that live close to the line.
View productWe take on a small number of partner builds each year, usually with operators who know a domain far better than we do.
Four products in active development. Nothing here is generally available yet, so every link below is a placeholder until launch.
Decision intelligence for operating teams.
View productAn operations copilot for work that happens in the physical world.
View productTurns what an organization knows into something it can be asked.
View productCash-flow intelligence for businesses that live close to the line.
View productFelixor exists because the interesting problem moved. It is no longer whether a model can do the thing. It is what to build now that it can.
We build products with AI at the core. Not products that added a chat panel, and not services that help someone else add one. Felixor conceives, designs, builds, and operates its own products, and takes a small number of partner builds each year with operators who bring a domain we could not learn fast enough on our own.
The studio model is a deliberate bet. AI products are unusually cheap to prototype and unusually expensive to get right, because the hard part arrives after the demo works: the evaluation, the failure modes, the trust, and the second session. A studio can carry a product through that stretch. A consultancy hands over a deck before it starts.
We stay small on purpose, we build in public where we can, and we would rather kill an idea in week three than defend it for a year.
Felixor was founded by Curtis Lee, who has spent his career on the user-facing edge of large technical platforms and on the small teams trying to build the next one.
Most recently he was a Vice President at Microsoft, where he led Azure Experiences and Ecosystems: the product surface of Azure, including its AI surface areas. It is an unusual vantage point on this moment. Azure is where an enormous amount of the world's AI capability is actually provisioned, and the job was to decide what that capability should look like to the people using it. Before that he ran Global Payments at Microsoft.
He has been a founder twice. Luxe, an on-demand parking and valet service, was acquired by Volvo. Pinwheel is a leading fintech API for direct deposit and bill switching. Earlier he worked in Corporate Development at Stripe, ran consumer products as a Vice President at Groupon, and held product roles at Google, YouTube, and Zynga.
The through line is consumer-grade product instinct applied inside infrastructure companies, which is more or less the exact job an AI product studio has to do.
Partner builds, early access to what is on the bench, or a conversation about something you cannot stop thinking about.
Every meaningful decision a company makes is argued once, decided once, and then forgotten. Quorum keeps the reasoning.
Teams do not lose decisions because nobody made them. They lose them because the context lives in a call nobody recorded, a thread nobody can find, and three people's heads. Six months later the same debate restarts from zero.
It listens to how a decision got made: the constraints, the tradeoffs, the option that was rejected and why.
When a team starts relitigating, Quorum puts the prior decision and its context in front of them first.
Every decision rests on assumptions. Quorum flags them when the world changes underneath.
This link goes live at launch. Until then, tell us what you would want it to do.
Software ate the office long before it touched the warehouse floor. Camber is built for the second one.
Operations teams run on exception handling: the shipment that missed, the route that broke, the crew that called out. That work is judgment under time pressure with incomplete information, and almost none of it is written down anywhere a model can learn from.
Camber works from the messy signal operators actually get: texts, calls, scans, and half-filled forms.
Every recommendation arrives with the constraints it weighed, so a dispatcher can overrule it in one glance.
Every operation has conventions that exist nowhere in the SOP. Camber picks them up from how people actually decide.
This link goes live at launch. Until then, tell us what you would want it to do.
Most companies do not have a knowledge problem. They have a retrieval problem that everyone has quietly given up on.
The answer usually exists. It is in a document from two reorgs ago, written by someone who left, using terminology the current team does not use. Search fails because the question and the answer do not share a single word.
Understory maps how your organization actually talks about its own work, including the terms that changed.
Every answer names its source, its date, and how confident it is that the source is still current.
The most valuable output of a knowledge system is an honest gap. Understory is built to report one.
This link goes live at launch. Until then, tell us what you would want it to do.
Most small businesses do not fail from bad margins. They fail from good margins and bad timing.
A profitable business can still run out of money on a Tuesday. The information needed to see it coming is spread across a bank, a processor, a payroll provider, and an accounting tool that reconciles once a month.
Tidemark forecasts forward from real transaction flow rather than reporting backward from the ledger.
A forecast is only useful if it points at a date and a cause. Tidemark does both.
It ranks the levers by how much they change the forecast, so the smallest useful action is obvious.
This link goes live at launch. Until then, tell us what you would want it to do.